# Gorbani Evaluation

## Self introduction

Talking about master thesis: worked with a control algorithm on a quadrotor

Simulation

Updated the firmware in C language

#### Erasmus

Gotenberg
Robust control of an inverse pendulum.

The IP han carry an object on top (parameter uncertainty)

## Questions

- Why do you need Geometric control on the quadrotor?

In literature there are 3 types (linear(PID LQR)/nolinear/geometric)

The linear uses the linearized model and it can become inaccurate with large angles. Also it uses Euler angles which can suffer from singularites (Gimble lock).

THe nonlinear solves the issue of linearization but still has an issue with the singularities..

The geometric control solved the issue of singularities

- What kind of simulator did you use?

Used a framework developed by a student in PoliMi. It could simulate a swarm of CrazyFlie

There was a cascade control so he modified it and implemented it

- How is the integration is done in the simulation - integration of the MASS matrix for instance..

The Simulator was in Gazebo, the models and plugins probably were the custom parts.

- How did you compare the control laws?

He didn't control the control algorithms but he compared the simulation and real robot.

When the angles roll/pitch were around 20 degs the results were similar.. But with higher values the results were different,, this was probably due to the difference in Mass matrix.

- What are the top two difficult things in your projects? and how you delt with them?

When implemented the controller in the real robot. he had to select the right gains.. around 12 values,, he spent around a week in tuning.

Also when he started learning ROS, he had a course in the begining, but the learning was a bit hard..

- Code?

In simulation he used Python but with the real robot he wrote the firmware in C.

- Alone?

He worked alone in the project

- Why did you apply to IIT?

He developed an interest in Contol systems,, and he heard about IIT and he became interested

- Interested in PhD?

Yes he would like to in the future..

- Other Projects

Design of a PID control of active suspension of a car

MATLAB/Simulink in simulation

- Have you done any projects with CAD software?

2 courses in bachelor in CAD (inventor and solid-works)

3rd year he used CATIA and inventor to make assemblies..

- Any parctical side with the CrazyFly?

He modified the firmware,,

USB radio to flash the firmware and send the trajectories..

- Do you have an experience with linux?

Yes

- Questions from the candidate?

He asked about the procedure..
Silvio explained

- Do you have a preference in getting feedback about the result.

He had another interview and he will get their reply soon (next Monaday and Tuesday).

Silvio explained that next week he can't have the answer. Mostly the week after, but he may at least get the result of the technical interview..

- Why did you minimize the (snap) instead of minimizing the jerk?

He found it in a paper and he copied it.

## Exercises

### Prob 1: Fan on inclined plane (10/10)

Started reading the text 11:34am
Done in 11:41

Since the velocity is constant the acceleration is 0
tension

Final result is correct according to silvio


### Prob 4: LTI stability I (8.5/10)

Started in 11:45
Ended 11:51

Candidate answer
1) xdot = 0; A.x = 0; x=0 if the rank is full only this solution
2) If the matrix is not full rank. Candidate says there are more than one solution. We have infinite solutions
3) real part of eigen values equal to negative for asymptotic statbility. Also answered correctly for zero real values

### Prob 7: Dice and coin (10/10)
Started 11:56
Ended 12:02

1) for the first point. The first dice is 4 and the probability is 1/6 for tail. For head, candidate listed the probabilities and sumed them. the Final answer (1/8)
2) The second part is (2/3) correct.

Has good reasoning and approach.


### Programming ex 7: Polynomial fitting (7/10)
Started: 12:09
Ended: 12:18 stuck

Candidate reasoning
1) Candidate identified Quadratic polynomial as a good fit. Doesn't remember the best fit method
2) Silvio gave some inputs about pseudoinverse
3) Candidate: err = y - (ax^2 + bx + c)
4) Candidate: has an idea about building matrix of x^2 x and 1 on one side and y vector on the other side of the equation

Restarted: 12:22
12:35 Cannot run the code on gitpod. But the code works.
Candidate hard-coded the coefficents while plotting.

Error while evaluating the coefficeints because the index was 0. Candidate initially did not focus on reading and understanding the error message.

Code

Notes

## Comments by Jury

### Silvio

### Affaf
- Has good reasoning and motivation for his thesis work. Worked alone on the thesis python code. Built up upon the firmware from another student.
- Candidate would like to take a PhD in the future.
- Active suspension PID for a car (another project). Used MATLAB and simulink.
- Has worked with ROS
- Overall seems calm and composed when explaining the project. Knows his work to a good level of depth

After the problem-solving step
1) Candidate has good reasoning and good logical steps to solve the problem. Has sound knowledge base
2) Candidate is receptive to the given inputs and hints. And uses them to proceed further.
3) Candidate: What is process after the interview? Candidate is waiting for another interview result which he will receive by Monday (16 Jan)

### Hosam

In general the evalaution is positive. He graduated last month and his master thesis is relevant and interesting. He has good understanding of the techinical details and he implemented it in both simulation and real robot.

His performance in the Exercises was really good because he got the right answers and followed good reasoning.

While explaining his projects he showed motivation and interest in controls and robotics.

I think he may proceed to the next step!